Negotiation, Politics, and the Work That Isn't in the Ticket

A substantial part of senior engineering work never appears in any tracked artifact. It's usually labelled politics, said with mild distaste, and treated as the tax paid for doing real work. It's also the part that isn't being automated, which makes it worth describing accurately rather than dismissively.

What the work actually is

Getting another team to do something. The change requires the platform team to expose a field. They have their own roadmap and no obligation. Making it happen is persuasion, timing, and reciprocity.

Deciding what not to do, and defending it. Saying no to a request from someone with standing, and having the no hold. This requires being trusted, which is accumulated over prior interactions.

Sequencing around other people's constraints. The migration must happen before the other team's rewrite, which means influencing their schedule, which means understanding their pressures.

Building the case for unfunded work. Technical debt, reliability investment, tooling. None of it is a feature; all of it needs someone to argue for it in terms decision-makers weigh.

Absorbing conflict. Two stakeholders want incompatible things. The engineer in the middle either resolves it or becomes the place the conflict lands.

→ None of this is written down as a task, none is estimated, and much of it happens in conversations that leave no artifact. It's invisible in every system that tracks work, which is why it reads as overhead.

Why it resists automation specifically

The inputs are relationships. Whether another team helps depends on history, trust, and what you've done for them. That's a state that lives between people.

The outputs aren't verifiable. There's no check for whether a negotiation went well. You find out over months, confounded by everything else.

Standing is required. The persuasive force of "I think we should do X" depends on who's saying it and their track record. An agent has no standing, and standing isn't transferable.

Much of it is reading what wasn't said. The hesitation, the topic avoided, the enthusiasm that's polite rather than real.

⚠️ Worth noting an uncomfortable corollary: work that resists automation because it depends on relationships and standing also tends to advantage people who are already well-connected. This isn't a meritocratic sorting, and it's worth naming rather than presenting the shift as neutral.

What changes as implementation gets cheap

The ratio shifts. If implementation was half the job and it compresses substantially, then coordination, negotiation, and persuasion become a larger fraction of what remains — not because there's more of it, but because there's less of everything else.

That has two implications:

It becomes more of the evaluation. If output volume is no longer differentiating, what distinguishes engineers is increasingly the ability to get the right things decided and the right work funded.

The people who dislike it are more exposed. Plenty of engineers chose the field partly to avoid this. That preference was accommodatable when deep implementation skill was scarce enough to be valuable on its own. It's less accommodatable as that scarcity falls — and saying so plainly is more useful than pretending otherwise.

✅ Getting better at it without becoming a different person

The framing that makes this tractable for people who find it distasteful: most of it is not manipulation, it's information flow and reliability.

• Understand the other side's constraints before asking. Most refusals are about their pressures, not about you. Asking what those are is both effective and honest.
• Make the ask specific and small. "Can you expose this field in the response?" gets answered. "Can we talk about the integration?" gets deferred.
• Be reliable. The largest input to whether people help you is whether you delivered last time. That's not politics; it's track record.
• Argue in their terms. Technical debt framed as engineering hygiene loses. Framed as "this is why the last three estimates slipped" it lands, because it's now about something the listener already cares about.
• Write the decision down afterward. Converts a negotiated outcome into something that persists past everyone's memory, which is also how you avoid re-negotiating it.

The takeaway

The unticketed work — persuading, sequencing, defending, absorbing conflict — is real work, it's a growing share of the job as implementation compresses, and it isn't being automated because its inputs are relationships and its outputs aren't checkable. Most of it is information flow and reliability rather than manipulation, which makes it learnable by people who'd rather not think of themselves as political. And the ratio is shifting whether or not anyone likes it.